Researchers from INTA Concepción del Uruguay evaluated a methodology that combines Sentinel-2 satellite images with field biomass measurements to estimate pasture availability on livestock farms. They analyzed 68 spectral indices and found that the models require periodic recalibration to maintain their accuracy.
Knowing how much pasture is available on a farm is a central variable for organizing the herd, defining stocking rates and anticipating possible forage deficits. With that goal, researchers from INTA Concepción del Uruguay are evaluating a methodology that combines Sentinel-2 satellite images, field biomass measurements and data from each farm to estimate pasture stock over large areas without having to walk and measure every paddock.
For producers, that information has concrete consequences: it allows them to decide how many animals to keep, when to move the cattle, how long to keep them in a paddock and how to manage reserves during periods of lower growth. The problem is that taking frequent measurements in every paddock requires time and labor, especially on large farms with heterogeneous environments.
According to Martín Durante, a researcher at INTA Concepción del Uruguay, the development integrates three sources of information: "Sentinel-2 satellite images, biomass measurements taken at selected points in the field and knowledge of the characteristics and management of each establishment." Spectral indices are calculated from the images — indicators that analyze how vegetation responds to light at different wavelengths; that response is related to vegetation cover and, through mathematical models, is used to estimate biomass.
For the satellite to translate that information into kilos of pasture, it first needs references taken directly in the field. Researchers measure biomass at selected points that serve as calibration sites, relate those data to the spectral response recorded by the satellite and adjust models that extend the estimates to the rest of the farm. "Satellite information provided spatial continuity and made it possible to weight the area occupied by each environment; field measurements turned the spectral response into quantitative biomass," Durante explained.
To evaluate the methodology, the team analyzed 68 spectral indices in temperate pastures of the Unidad Integrada Balcarce, made up of INTA and the Universidad Nacional de Mar del Plata, based on records from 46 sampling dates. The research also explored its application in natural grasslands of the Río de la Plata region, with 11 sampling dates at five sites in Argentina and Uruguay, including farms in Concepción del Uruguay and Gualeguay, in Entre Ríos. The work involved researchers from Argentina, Uruguay, Colombia and Costa Rica, within the framework of a FONTAGRO project.
Researchers identified a limitation to be resolved before widespread use: the models lost accuracy when applied to images from dates later than those used for their calibration. That is why they raise the need for periodic recalibrations so that estimates remain reliable. As part of the development, they built a pilot web application that integrates Sentinel-2 spectral indices with biomass measurements to generate pasture availability maps at farm scale. The advances will be presented during Caravana Pampa 2026, a regional initiative that will tour farms in Brazil, Uruguay and Argentina to exchange experiences on livestock production and conservation of natural grasslands.
