Soft Commodity Infrastructure (Brazil)
Where Brazilian farms connect to markets. This is the geolocated midstream of Brazilian agriculture: the silos, slaughterhouses, dairy plants, and sugar/ethanol mills that farms sell into, plus the cooperatives that farms belong to. It exists to answer, for any rural property, who to contact about deforestation on the land — the cooperative that has the farmer as a member, or the buyer that receives their grain, cattle, or cane.
How it was made
Every point comes from a government registry, or is geocoded from one. Nothing is guessed from satellite imagery.
- Grain silos — CONAB's national warehouse registry (
ArmazensCadastrados), which every grain warehouse must register with. It ships real GPS coordinates, so these are exact. Includes cooperative, trader/private, and official silos. - Slaughterhouses & dairy plants, federal — MAPA's SIF (federal meat/dairy inspection) registry. Has the company, tax ID, and address but no coordinates, so we geocoded the addresses ourselves. Beef, poultry, pork, dairy.
- Slaughterhouses & dairy plants, state/municipal — the smaller local plants (SIE/SIM/CONSORCIO) that a frontier producer often delivers to, which the federal list misses. These live in MAPA's SISBI system, which has no bulk download; Trase pulled the SISBI records and geocoded them from the registry addresses, and we use that subset. Registry-derived, not AI-detected.
- Sugar/ethanol mills — ANP's ethanol producer list (company, tax ID, town, capacity). No street address, so each mill sits at its town's center.
- Cooperative membership — IBGE's Censo Agropecuário 2017, which counts how many farms in each município belong to a cooperative. A density signal, not a named cooperative — but it works even in remote frontier towns.
- Municipality geometry — IBGE's official município boundaries.
Trust labels on every row
Because not every point is equally precise, each carries two labels. basis
is observed (someone official recorded it) or modelled (we calculated it — only
the gravity-catchment rows). geom_method is registered (the source gave us
the GPS coordinate — CONAB silos), geocoded (we turned an address into a point —
can be off), or municipality_centroid (all we had was the town). A user can see
at a glance how good each location is.
Full methodology & source
Pipeline, tests, and a plain-language writeup live at
https://github.com/wri/rural-land (see soft_commodity_infrastructure/METHODOLOGY.md).
Brazil only. Rice, cotton, and coffee are not covered — no authoritative open
geolocated source exists for them.
Collections
facilities
Facility and membership layers for Brazilian soft-commodity supply chains. One unified layer plus the components it's built from. Geometry is WGS84 (EPSG:4326).
The files
BR_facilities...
Coverage
Spatial Extent
- West: -73.9904, South: -33.7508, East: -29.2990, North: 5.2718
Source
https://github.com/wri/rural-land
Processing Notes
Facilities compiled from government registries. Grain silos: CONAB national warehouse registry (registered coordinates), cooperative + trader + official. Slaughterhouses & dairy plants: MAPA SIF (federal, address-geocoded) plus Trase SISBI (state/municipal, registry-geocoded). Sugar/ethanol mills: ANP producers, municipality-centroid. Cooperative membership: IBGE Censo Agropecuário 2017 by município. Every geometry carries a basis (observed | modelled) and geom_method (registered | geocoded | municipality_centroid) so precision is explicit. No AI-detected geometry. BR_facilities.parquet is the unified layer (all tiers); the other files are the component layers. See METHODOLOGY.md in the source repo for full detail.
Attribution
Soft Commodity Infrastructure (Brazil), WRI. Sources: CONAB, MAPA/SIF, ANP, IBGE, Trase (SISBI) — CC-BY 4.0.
License
Contact
Tristan Grupp (WRI) tristan.grupp@gmail.com
Generated by Portolan from STAC metadata and .portolan/metadata.yaml