← Journal

Compliance

EUDR for Indonesian plantations: what the polygon requirement means

The EU Deforestation Regulation applies from 30 December 2026. For palm oil, rubber, coffee, and cocoa suppliers, the hardest part is spatial: a clean, accurate polygon for every plot. Here's what that involves.

This article explains the spatial side of the EUDR for GIS teams. It isn’t legal advice. Check your obligations with your buyers and a qualified adviser.

The regulation in one paragraph

The EU Deforestation Regulation (Regulation (EU) 2023/1115, “EUDR”) says that cattle, cocoa, coffee, oil palm, rubber, soya, and wood, and many products made from them, can only be sold in or exported from the EU if they are:

  1. deforestation-free: produced on land that was not deforested after 31 December 2020;
  2. legal: produced in line with the laws of the country of production;
  3. covered by a due diligence statement submitted by the company placing them on the EU market.

When it applies

After two postponements, the rules now apply from:

  • 30 December 2026 for large and medium-sized operators and traders;
  • 30 June 2027 for micro and small operators.

The European Commission’s simplification review in May 2026 ruled out another delay. If you supply EU buyers, expect requests for plot data well before those dates.

Why this is a GIS problem

The due diligence statement must include the geolocation of every plot of land where the commodity was produced:

  • coordinates with six decimal places (roughly 10 cm precision);
  • for plots larger than 4 hectares (for commodities other than cattle), a polygon of the plot boundary;
  • for smaller plots, a single point is accepted.

For a company sourcing from thousands of smallholders, that means a plot inventory of thousands of points and polygons that has to be accurate, current, and defensible.

What usually goes wrong

From a GIS point of view, most problems are familiar:

  • Invalid geometries: self-intersections, unclosed rings, and duplicate vertices that break area calculations.
  • Overlapping plots: two farmers’ polygons covering the same land, or plots overlapping a concession boundary.
  • Wrong coordinate system: data collected in UTM or a local datum, then exported as if it were latitude/longitude.
  • Points where polygons are required: plots recorded as a single GPS point even though they’re larger than 4 ha.
  • Polygons that don’t match reality: boundaries walked years ago, or digitized from old imagery.
  • Missing links: plots that can’t be traced to the deliveries that came from them.

A practical workflow

  1. Build the plot inventory. Collect boundaries from field teams (QField, Survey123, KoboToolbox), existing concession maps, and smallholder registries.
  2. Clean the geometry. Fix invalid shapes, resolve overlaps, and convert everything to WGS 84 latitude/longitude with six decimals.
  3. Apply the 4 ha rule. Calculate area in a suitable projected CRS, flag plots over 4 ha that only have a point, and plan field visits to capture their polygons.
  4. Check against forest cover. Compare each plot with a forest baseline around 31 December 2020 and with later imagery to flag possible clearing. Reference maps such as the EU’s Global Forest Cover 2020 help, but they’re not legally binding, so flagged plots need a closer look.
  5. Investigate flags. Use higher-resolution imagery or field visits to confirm whether a flagged change is deforestation, replanting, or a mapping error.
  6. Keep a record. Store the version of each polygon, the imagery used, and the result of each check, so you can answer buyer and auditor questions later.

What we’re exploring at Yerin

This workflow is mostly repetitive spatial checks, which is the kind of work an AI assistant for GIS should handle: cleaning plot geometries, checking them against forest baselines, flagging changes, and preparing the geolocation data in the format buyers ask for. It’s on our Labs list as an idea we’re researching, and we’d like to hear from teams facing it now.

Sources

Keep reading

Free vs commercial satellite imagery: when 30 cm is worth paying for

Sentinel-2 and Landsat cover most monitoring work at no cost. Here's when commercial imagery from Airbus, Planet, or Vantor (formerly Maxar) earns its price, what it costs in 2026, and the licensing trap to avoid.

Read the article →

GeoLibre review: a GIS that runs in your browser

A free, open-source, cloud-native GIS that loads, edits, and analyzes spatial data on your device. What it does well, where QGIS still wins, and when we'd use it.

Read the article →