Modelos de predicción de biomasa en sistemas de producción lechera usando índices de vegetación de drones e imágenes satelitales

Accurate estimation of pasture biomass is essential for decision-making in grazing systems, it requires information obtained at different spatial scales using diverse measurement methods. This study integrates multiple datasets generated from experiments conducted at the Centro Regional Sur (CRS) Research Center, combining direct field measurements, such as: sward height measured with a ruler, compressed height obtained using a rising plate meter (RPM), and C-Dax pasture meter, using the Normalized Difference Vegetation Index (NDVI) derived from drone and satellite imagery. The goal was studying the relationships between structural and spectral vegetation variables and performance evaluation of different predictive models of pasture biomass.

Results revealed consistent and predominantly non-linear associations between direct measurements and NDVI, particularly for drone-derived NDVI, where patterns related to index saturation and differences in canopy architecture among species were observed. Models based on direct field measurements showed higher explanatory power at the fixed-effects level; however, models built using drone-derived NDVI achieved comparable predictive performance during validation. In addition, the satellite-based NDVI model exhibited greater stability and performance comparable to or better than the drone-based model, likely due to its broader spatial scale, which attenuates fine-scale within-paddock heterogeneity.

Overall, these findings highlight the complementarity between in situ measurements and remote sensing data, and support the use of integrated approaches for pasture biomass prediction. Such integration balances accuracy, applicability across different spatial scales, and practical implementation, offering strong potential to improve grazing management decisions and production planning.