Satellite monitoring of quantity and quality of available biomass in pastoral livestock systems

Three key tools were developed to advance the satellite monitoring of forage biomass quantity and quality in grasslands: a field sampling protocol, a mobile application for recording field observations (biomass and forage quality), and a web platform for visualizing collected data and supporting sampling activities. These tools enabled the establishment of a monitoring network with the active participation of more than 79 researchers and technicians. The database currently contains approximately 3,600 records (1,600 collected directly under the standardized protocol and 2,000 obtained through indirect methods). Using these data, machine learning models (Random Forest) were evaluated with Sentinel-1 and Sentinel-2 satellite imagery. The relationship between biomass and satellite-derived variables showed strong spatial and temporal variability, resulting in paddock-scale prediction errors (RMSE) of 820 and 612 kg DM/ha (~41% and 32%) when considering all forage resources and temperate pastures from Balcarce, respectively. In contrast, the relationship between satellite data and forage quality (crude protein content) proved to be more consistent and generalizable. These findings supported the development of two simplification tools that integrate satellite and field data to improve estimation accuracy. The first is a spatial simplification approach that optimizes sampling through measurements at strategically selected locations (corresponding to extreme spectral index values). The second is a temporal simplification approach that estimates real-time forage stocks by integrating residual biomass measurements with forage growth models. Both tools were validated under real farming conditions, achieving operational errors of approximately 20%, and contribute to sustainable livestock management based on scientific evidence. Dissemination and training activities reached more than 3,500 participants.