INPE deforestation data: what PRODES and DETER can and cannot prove

Two systems, two different jobs

Brazil’s National Institute for Space Research, INPE, runs two well-known deforestation monitoring systems, and conflating them is the most common error in this area.

PRODES is the annual consolidated mapping of clear-cut deforestation. It is careful, reviewed, uses higher-resolution imagery, and produces the official yearly figures. It is the number quoted in national statistics and international reporting.

DETER is a near-real-time alert system. It sweeps frequently at lower resolution to flag likely change quickly, so that enforcement can respond while it matters. It is explicitly designed as a targeting tool for enforcement, not as a measurement system.

Why the difference matters to a buyer

A DETER alert says a satellite pass detected something consistent with forest change in an area, quickly, at lower confidence. It may be confirmed later by PRODES, or it may not — cloud, shadow, seasonal change and selective logging all complicate the picture.

Presenting a DETER alert as confirmed deforestation attributable to a supplier is a serious overstatement, and it is one we see made. We surface DETER separately from PRODES, and describe it as preliminary, for exactly this reason. If a vendor shows you deforestation alerts without telling you which system produced them, ask.

Satellites see land, not companies

This is the structural limitation, and no amount of data quality removes it. Both systems produce mapped polygons: areas where forest cover changed. Neither carries a company name, a tax identifier, or an owner.

To connect a polygon to a supplier you need a property boundary to overlay it on, and in Brazil that means the rural environmental registry, CAR. Which in turn means you need your supplier’s CAR code — and CAR boundaries are self-declared by the landholder and not independently verified by the registry.

So the chain from satellite to supplier runs: alert or mapping, overlaid on a self-declared boundary, supplied by the party being checked. Every link there is real and useful. None of them is proof, and the last one deserves particular attention.

And deforestation is not automatically illegal

Brazilian law permits legal clearing under certain conditions and with the right authorisation. Detected forest loss on a property is therefore not, in itself, evidence of a breach of Brazilian law — which matters directly under the EU Deforestation Regulation, because that rule asks two separate questions: whether production was deforestation-free after the cut-off date, and whether it complied with the producing country’s law.

A finding can be relevant to the first question and still be entirely lawful under the second. Treating detected clearing as automatic wrongdoing is unfair to the supplier and weakens your own file, because it is an inference you cannot support.

What to do with it in practice

Use it to prioritise, not to conclude. Satellite evidence is excellent for deciding which of forty suppliers deserves a real conversation and which does not. It is poor as the sole basis for terminating one.

Where you have a CAR code, an overlay result with dates and the system that produced it is a legitimate part of a due-diligence file — recorded with its limitations, including that the boundary was self-declared. Where you do not have a CAR code, be honest in your file that the geographic check could not be performed, rather than leaving a gap that reads as a clean result.

Enforcement records are the complement. Where IBAMA has actually acted, there is an embargo record naming a party — an authority’s finding rather than an inference from imagery.

Common questions