Brazil is the world’s third-largest corn producer after the United States and China. Corn production is concentrated in four of Brazil’s 26 states: Mato Grosso, Minas Gerais, Goias, and Parana.
Gro’s Brazil Corn Forecast Yield Model is updated daily, providing in-season yields at the district, province, and national levels throughout the growing season.
When combined with our other Corn Forecast Yield Models, for China, the United States, and Argentina, Gro is able to forecast more than 70% of global corn production in a given year, providing users with a highly accurate, real-time window into global corn supply.
How the Model Was Built
Gro’s district-level yield model for corn in Brazil was built using machine-learning techniques that rely on weather and environmental data. The model performs particularly well in high yield and production areas.
Modeling corn yield in Brazil presents particular challenges when compared with other yield models we have built at Gro. Brazil has two corn growing seasons per year, making it harder to model annual production. The breakdown of the two crops has also shifted: The second corn season, known as the safrinha crop, has become the greater producing period since 2012 as soybeans have taken priority over corn during the summer season, when the first corn crop is cultivated.
Crop production also varies temporally and spatially. There are significant differences in weather, soil conditions, and technology adoption throughout the country, which results in high spatial variation of crop yields. As a result, corn yield is higher in central and southern Brazil, and lower in the north.
Gro’s Brazil Corn Forecast Yield Model uses the following variables:
- Gro Crop Cover
- Yield, area and production history in Brazil
- 8-day normalized difference vegetation index (NDVI)
- Land surface temperature
- Weather forecasts of air temperature and precipitation
- Evapotranspiration difference from 10-yr median (2003-2013)
- Drought severity and coverage index
- Soil characteristics
- Latitude and longitude
- 30-meter resolution cropland masks